there is no true count. a company in stealth has not registered itself anywhere countable, which is the point of being in stealth.
so every number you will see measures detection, not reality. here are the three that exist and what each one is actually counting.
the self-declared number: ~40,000
roughly forty thousand people list a "stealth startup" style employer on linkedin at any given time, with about 105 entering and 114 leaving that state daily.
what it counts: people who chose to label themselves. that includes founders. it also includes people between jobs, consultants, and anyone who set the field once and never updated it. it excludes every founder who left their previous title in place, which is common, especially for people who have not told their old employer.
treat it as a floor on a self-selecting population, not a census.
the judged number: 896
frontrun has flagged 2,549 people building in stealth and rates 896 of them strong.
different measurement entirely. we do not enumerate a linkedin field. three detectors run over the investor follow graph: a bio changing into founder language, a move into a stealth or unnamed role, and tracked investors converging on a person with no company. each hit is then judged against verified career evidence and returned as strong, worth a look, or weak.
the gap between 2,549 and 896 is the filtering. 64% do not survive it, which is roughly the false-positive rate you should assume for any unfiltered stealth list.
the number that is not a trend
our monthly detection volume:
| month | people flagged | rated strong |
|---|---|---|
| july 2026 | 322 | 139 |
| august 2026 | 416 | 166 |
| september 2026 (to the 17th) | 1,819 | 591 |
that jump is our pipeline expanding, not stealth activity tripling. we added detectors and widened the graph during that window.
we are stating it because the chart is tempting and the honest reading is boring. if a vendor shows you a rising stealth-activity curve without telling you whether their own coverage changed, the chart is measuring the vendor.
where they come from
the more durable statistic is composition. across 143 previous employers with at least three founders each:
| previous company | founders flagged in stealth |
|---|---|
| Meta | 69 |
| 68 | |
| Microsoft | 65 |
| Amazon | 47 |
| Apple | 41 |
| OpenAI | 26 |
| Uber | 19 |
| Tesla | 18 |
| Google DeepMind | 18 |
| Coinbase | 17 |
the big five dominate on headcount alone. the more interesting entries are the ones punching above their size: openai at 26 from a company of a few thousand is a far higher rate than microsoft at 65 from a company of 220,000.
full list of all 143 at founders building in stealth, by where they came from.
what any of these numbers are good for
not much on their own. a count tells you the category is real. it does not tell you which person to email.
the number that changes a decision is a dated flag on one person, because it can be checked afterwards. jacob zietek was published on a public list on august 19 2026 and announced a pre-seed from a16z and accel 21 days later. that is one row, and it is worth more than any aggregate on this page.
related
open source deal flow
companies and founders surfaced on frontrun. curated, every week.
